CN-121982083-A - Image registration method and device

CN121982083ACN 121982083 ACN121982083 ACN 121982083ACN-121982083-A

Abstract

The application relates to an image registration method and device, which are applied to an autonomous operation system, and the method comprises the steps of acquiring a target key frame image, an acquired current frame image and a mask image of an interested region in the target key frame image; the method comprises the steps of determining a first target deformation field according to a current frame image, a target key frame image and a mask image, performing deformation processing on the target key frame image according to the first target deformation field to obtain an interested region in the current frame image, updating the target key frame image according to the acquired current frame image, and returning to execute the step of determining the first target deformation field according to the current frame image, the target key frame image and the mask image according to the updated target key frame image. The image registration method provided by the application has higher registration precision.

Inventors

OUYANG XIAOYUN
NI HONG
An Sile
YANG FEI

Assignees

武汉联影智融医疗科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (10)

1. A method of image registration for application to an autonomous surgical system, the method comprising: Acquiring a target key frame image, an acquired current frame image and a mask image of an interested region in the target key frame image; Determining a first target deformation field according to the current frame image, the target key frame image and the mask image; performing deformation processing on the target key frame image according to the first target deformation field to obtain an interested region in the current frame image; updating the target key frame image according to the acquired current frame image, and returning to execute the step of determining a first target deformation field according to the current frame image, the target key frame image and the mask image according to the updated target key frame image.
2. The method of claim 1, wherein updating the target key frame image from the acquired current frame image comprises: Determining a candidate frame image according to the acquired current frame image; and updating the target key frame image according to the candidate frame image.
3. The method of claim 2, wherein the first target deformation field comprises a positive deformation field and an inverse deformation field, the determining a candidate frame image from the acquired current frame image comprising: Determining whether the current frame image meets a preset condition or not, wherein the preset condition comprises that the similarity between the current frame image and the target key frame image reaches a preset similarity threshold, the deformation field formed by compositing the positive deformation field and the inverse deformation field reaches a preset deformation field, and the similarity coefficient between the mask image and a new mask image reaches at least one of preset coefficient thresholds; And under the condition that the current frame image meets the preset condition, determining the current frame image as the candidate frame image.
4. A method according to claim 3, wherein said updating said target key frame image from said candidate frame image comprises: Determining whether the acquisition interval between the target key frame image and the current frame image reaches a preset interval threshold value; determining whether the target key frame image meets the preset condition or not under the condition that the acquisition interval reaches the preset interval threshold; And determining the candidate frame image as a new target key frame image under the condition that the target key frame image does not meet the preset condition.
5. The method of claim 1, wherein the determining a first target deformation field from the current frame image, the target key frame image, and the mask image comprises: And inputting the current frame image, the target key frame image and the mask image into a pre-trained registration network to obtain the first target deformation field.
6. The method of claim 5, wherein the training process of the registration network comprises: The method comprises the steps of obtaining a training sample, wherein the training sample comprises a current frame image sample, a key frame image sample and a mask image sample of an interested region in the key frame image sample; and training an initial network model according to the training sample to obtain the registration network.
7. The method of any one of claims 1-6, wherein the target key frame image is a target preoperative image and an intra-operative image and the current frame image is an intra-operative image.
8. The method of claim 1, wherein the current frame image is an intra-operative image and the target key frame image is a target pre-operative image, the method further comprising: Determining a second target deformation field according to the intraoperative image, the target preoperative image and the mask image corresponding to the target preoperative image; And carrying out deformation processing on the target preoperative image according to the second target deformation field to obtain a region of interest in the intra-operative image.
9. The method of claim 8, wherein the method of acquiring the target preoperative image comprises: Acquiring an initial preoperative image and an initial mask image corresponding to the initial preoperative image; And carrying out rigid registration on the initial preoperative image and the intra-operative image to obtain the target preoperative image and a mask image corresponding to the target preoperative image.
10. An image registration apparatus for use in an autonomous surgical system, the apparatus comprising: the acquisition module is used for acquiring a target key frame image, an acquired current frame image and a mask image of an interested region in the target key frame image; the determining module is used for determining a first target deformation field according to the current frame image, the target key frame image and the mask image; the processing module is used for carrying out deformation processing on the target key frame image according to the first target deformation field to obtain an interested region in the current frame image; And the updating module is used for updating the target key frame image according to the acquired current frame image, and returning to execute the step of determining a first target deformation field according to the current frame image, the target key frame image and the mask image according to the updated target key frame image.

Description

Image registration method and device Technical Field The present application relates to the field of image processing technologies, and in particular, to an image registration method and apparatus. Background In clinical medicine, surgery using a surgical robot system is widely used, for example, puncture surgery is performed by a puncture surgical robot system using a master-slave puncture or off-line puncture technique. The core task of the surgical path is to map the target point of the preoperative planning into an intraoperative image continuously acquired in the operation and drive a surgical tool in the surgical robot to reach the target point. In the conventional technology, a preoperative image and a continuously acquired intra-operative image are usually registered directly, so that a planned target point in the preoperative image is mapped into the intra-operative image, so that a doctor can perform an operation according to the target point in the intra-operative image. However, the registration method in the conventional art has a problem of low registration accuracy. Disclosure of Invention In view of the foregoing, it is desirable to provide an image registration method and apparatus capable of improving registration accuracy. In a first aspect, the present application provides an image registration method for use in an autonomous surgical system, the method comprising: Acquiring a target key frame image, an acquired current frame image and a mask image of an interested region in the target key frame image; Determining a first target deformation field according to the current frame image, the target key frame image and the mask image; Performing deformation processing on the target key frame image according to the first target deformation field to obtain an interested region in the current frame image; Updating the target key frame image according to the acquired current frame image, and returning to execute the step of determining the first target deformation field according to the current frame image, the target key frame image and the mask image according to the updated target key frame image. In one embodiment, updating the target key frame image according to the acquired current frame image includes: Determining a candidate frame image according to the acquired current frame image; and updating the target key frame image according to the candidate frame image. In one embodiment, the first target deformation field comprises a positive deformation field and an inverse deformation field, and determining the candidate frame image from the acquired current frame image comprises: Determining whether the current frame image meets a preset condition or not, wherein the preset condition comprises that the similarity between the current frame image and the target key frame image reaches a preset similarity threshold value, the deformation field formed by compositing the positive deformation field and the inverse deformation field reaches a preset deformation field, and the similarity coefficient between the mask image and a new mask image reaches at least one of preset coefficient threshold values; and determining the current frame image as a candidate frame image under the condition that the current frame image meets the preset condition. In one embodiment, updating the target key frame image from the candidate frame image includes: determining whether the acquisition interval between the target key frame image and the current frame image reaches a preset interval threshold value; Under the condition that the acquisition interval reaches a preset interval threshold value, determining whether the target key frame image meets a preset condition or not; And determining the candidate frame image as a new target key frame image in the case that the target key frame image does not meet the preset condition. In one embodiment, determining the first target deformation field from the current frame image, the target key frame image, and the mask image comprises: And inputting the current frame image, the target key frame image and the mask image into a pre-trained registration network to obtain a first target deformation field. In one embodiment, the training process of the registration network includes: The method comprises the steps of obtaining a training sample, wherein the training sample comprises a current frame image sample, a key frame image sample and a mask image sample of an interested region in the key frame image sample; And training the initial network model according to the training sample to obtain the registration network. In one embodiment, the target key frame image is a target preoperative image and an intra-operative image, and the current frame image is an intra-operative image. In one embodiment, the current frame image is an intra-operative image, the target key frame image is a target pre-operative image, and the method further comprises: determining a second target deformation fi